Trade Theory and Trade Facts∗

Home , International economics, New trade theory

Federal Reserve Bank of Minneapolis Research Department Staﬀ Report 284

March 2001 (First version October 1999)

Raphael Bergoeing ILADES-Georgetown University, Universidad Alberto Hurtado

Timothy J. Kehoe University of Minnesota and Federal Reserve Bank of Minneapolis

ABSTRACT

This paper quantitatively tests the “new trade theory” based on product differentiation, increasing returns, and imperfect competition. We employ a standard model, which allows both changes in the distribution of income among industrialized countries, emphasized by Helpman and Krugman (1985), and nonhomothetic preferences, emphasized by Markusen (1986), to effect trade directions and volumes. In addition, we generalize the model to allow changes in relative prices to have large effects. We test the model by calibrating it to 1990 data and then “backcasting” to 1961 to see what changes in crucial variables between 1961 and 1990 are predicted by the theory. The results show that, although the model is capable of explaining much of the increased concentration of trade among industrialized countries, it is not capable of explaining the enormous increase in the ratio of trade to income. Our analysis suggests that it is policy changes, rather than the elements emphasized in the new trade theory, that have been the most significant determinants of the increase in trade volume.

∗ c 2001, Raphael Bergoeing and Timothy J. Kehoe. This paper grew out of the Bergoeing’s 1996 Ph.D. thesis at° the University of Minnesota. The authors thank the McKnight Foundation and the National Science Foundation for ﬁnancial support. Useful comments and suggestions have been provided by participants in numerous seminars and conferences, especially Caroline Betts, Harold Cole, Barbara Craig, Patrick Kehoe, David Hummels, Robert Lucas, Moshe Syrquin, Nancy Stokey, and Jaume Ventura. The views expressed herein are those of the authors and not necessarily those of the Federal Reserve Bank of Minneapolis or the Federal Reserve System. 1. Introduction

This paper investigates the extent to which the “new trade theory” can quantitatively match some of the facts that it was designed to explain. It does so by calibrating a standard model,basedonMarkusen(1986), to 1990 data and then “backcasting” to 1961 to see what changes in crucial variables between 1961 and 1990 are predicted by the theory.

Thenewtradetheory,developedbyresearcherslikeHelpman(1981), Krugman (1979), and Lancaster (1980) in the late 1970s and 1980s, was motivated by the failure of more traditional theories to explain some of the most signiﬁcant facts about post World War II trade data. As Deardorﬀ (1984) and Helpman and Krugman (1985) explain, the new trade theory was designed to account for three major facts:

The ratio of trade to income has increased. • Trade has become more concentrated among industrialized countries. • Trade among industrialized countries has been largely intraindustry trade. •

Figure 1 presents evidence for the ﬁrst fact, showing how much faster world trade has increased than world income. To make comparisons easy in the ﬁgure, data on both world trade volume, measured by summing up exports throughout the world, and world income have been expressed as indices where 1950=100. Over the period 1950-1990, the ratio of trade to income worldwide increases by 86.1 percent. The data for both trade and income used to derive the indices in Figure 1 are measured in constant 1970 U.S. dollars. Alternatively, we could look at trade as a fraction of income, dividing the current value of trade by the current value of income. As a fraction of the value of GDP, the value of trade increased from 7.9 percent in 1950 to 15.4 percent in 1990, a 94.9 percent increase. To make the second fact precise, we identify industrialized countries with the Organization for Economic Cooperation and Development (OECD), which was formed in 1961. Figure 2 shows how much faster trade within the OECD has increased than OECD trade with the rest of the world. The ratio of OECD-OECD trade to OECD-RW trade went from 0.84 in 1961 to 1.58 in 1990.

Evidence for the third fact can be found in the high Grubel-Lloyd indices of international trade in industrialized countries. For this sort of data it is more diﬃcult to calculate long time series. In 1990 though, Grubel-Lloyd indices based on two-digit SITC data from the

OECD say that 68.4 percent of OECD-OECD trade was intraindustry compared with only

38.1 percent of OECD-RW trade.

Closely related to the ﬁrst two facts is yet a fourth fact: The ratio of trade to income withintheOECDincreasedevenfasterthantheratiooftradetoincomeworldwide. Figure3 presents the relevant data. Trade within the OECD went from 5.3 percent of OECD income in 1961 to 11.2 percent in 1990. The ratio of trade to income within the OECD increased by

111.5 percent. By comparison the United Nations data say that the ratio of trade to income worldwide increased by only 59.3 percent over 1961-1990.

That the new trade theory was developed to explain these facts is explicit in textbook expositions by the developers of the theory. Helpman and Krugman (1985), for example, point out that conventional trade models like the Ricardian model and the Heckscher-Ohlin model cannot hope to explain these facts and go on to say,

These ... empirical weaknesses of conventional trade theory ... become under-

standable once economies of scale and imperfect competition are introduced into

our analysis.

2 Helpman and Krugman stress the changes in the distribution of income among industrialized countries as a major cause of the expansion of trade relative to income. In the early post war period the United States accounted for much of the world’s income and consumption.

As the distribution of national income became more equal, their model predicts that trade volumes should rise.

Focusing on the increase in trade among industrialized countries, Markusen (1986) stresses unequal income elasticity of demands that result from nonhomothetic preferences.

If demand for diﬀerentiated products is superior to that for homogeneous products, then intraindustry trade should expand as income rises; and, if industrialized countries are net ex- porters of these diﬀerentiated products, then intraindustry trade among industrialized countries should expand faster than other trade. As Markusen et al. (1995) point out, it was to match the facts listed above that the theory had been formulated:

Thus, nonhomogeneous demand leads to a decrease in North-South trade and to

an increase in [intraindustry trade] among the northern industrialized countries.

These are precisely the facts there were to be explained.

Our model generalizes those developed by Helpman and Krugman (1985) and Markusen

(1986) in that it allows changes in relative prices to have large eﬀects on trade volumes. Be- cause of faster total factor productivity growth in the manufacturing sector, the relative prices of manufactured goods have fallen sharply from 1961 to 1990 compared to the prices of primary goods and services.

Our numerical experiments show that, although the model can explain the increased concentration of trade among industrialized countries, it is not capable of explaining the

3 enormous increase in the ratio of trade to income. The simple fact is that it has been the trade of manufactured goods among OECD countries that accounts for most of the expansion of trade over the period 1961-1990. Over this same period, however, the consumption of manufactures in these countries has gone down as a fraction of income. A model that relies on the taste for variety approach developed by Spence (1976) and Dixit and Stiglitz (1977) links increases in trade to increases in consumption. It seems that it is policy changes, rather than the elements emphasized in the new trade theory, that have been the most signiﬁcant determinants of the increase in trade volume.

In related literature, Hunter (1991) estimates that nonhomothetic preferences accounts for about one fourth of observed interindustry trade. Helpman (1987) reports regression results that he interprets as support for the Helpman-Krugman explanation of the distribution of national income as the driving force behind trade increases. Hummels and Levinsohn

(1995), however, argue that Helpman’s results are not related to the Helpman-Krugman theory. Haveman and Hummels (1999) argue that the new trade theory models rely on taste for variety that is not consistent with the data and predict too much trade. Yi (1999) argues that the new trade theory models cannot account for the increase in trade unless they incorporate changes in both trade policy and international vertical integration.

It is worth noting that lack of data on trade in services in 1961 forces us to restrict our attention to merchandise trade in both our data analysis and in our theory. Information on sources for all of the data presented in this paper are included in the data appendix.

4 2. A “New Trade Theory” Model

Consider a world in which there are n developed countries, identiﬁed with the 23 countries in the Organization for Economic Cooperation and Development (OECD) in 1990, and a rest of the world. (There were actually 24 countries in the OECD in 1990, but, since data for Belgium and Luxembourg are aggregated together, we treat them as one country.)

Table 1

OECD in 1990 Country Share of Income (percent) Australia 1.3209 Austria 0.7153 Belgium/Luxembourg 0.9254 Canada 2.5475 Denmark 0.5791 Finland 0.6046 France 5.3609 Germany 7.3549 Greece 0.3718 Iceland 0.0280 Ireland 0.2042 Italy 4.9059 Japan 13.3195 Netherlands 1.2721 New Zealand 0.1933 Norway 0.5177 Portugal 0.3027 Spain 2.2189 Sweden 1.0304 Switzerland 0.1274 Turkey 0.6757 United Kingdom 4.3747 United States 24.9076

In each country or region, there are three types of goods, a primary good that is tradable and homogeneous, manufactured goods that are tradable and diﬀerentiated by the

ﬁrm that produces them, and a service good that is nontradable and homogeneous within the

5 country where it is produced. The OECD and the rest of the world diﬀer in the endowments of physical capital and human capital held by consumers. Speciﬁcally, the endowments of

OECD consumers are koe and hoe while those of consumers in the rest of the world are krw and hrw.

An individual consumer in country or region j, j =1,...,n,oe,rw, solves the problem of maximizing

η/ρ j η j ρ j η βp xp + γp + βm xm (z) dz + βs xs + γs 1 η (1) w − " µZD ¶ # Á ¡ ¢ ¡ ¢ subject to

j j j j j j j j j j j qpxp + qm (z) xm (z) dz + qsxs r k + w h xp,xm (z) ,xs 0. (2) w ≤ ≥ ZD

j j Here xp is the consumption of the primary good and qp is its price; xm (z) is the consumption

j of the manufactured good produced by ﬁrm z and qm (z) is its price; xs is the consumption

j j j of the service good and qs is its price; and r is the return to physical capital while w is the return to human capital. Notice that, since we assume that consumers in diﬀerent OECD

oe oe 1 2 n oe countries have the same endowments k and h , xi = xi = ... = xi = xi . The parameter

ρ, 1 ρ > 0, governs the elasticity of substitution 1/ (1 ρ) between any two differentiated ≥ − manufactured goods in the interval Dw =[0,dw] of such goods produced throughout the world; the parameters γp and γs govern the income elasticities of demand for the different types of goods; and the parameter η governs the elasticity of substitution between any two types of goods, which in turn governs the price elasticities of demand for the different types

6 of goods. In the base case, where γp = γs = η =0, all of the income elasticities and price elasticities are equal to one. In this case, the utility function is

1/ρ j j ρ j βp log xp + βm xm (z) dz + βs log xs. (3) w µZD ¶

The population of each country or region j, j =1,...,n,oe,rw is N j.Ofcourse,we require n N oe = N j. (4) j=1 X The aggregate endowments of human and physical capital are respectively

Hj = N jhj (5)

Kj = N jkj. (6)

In the homothetic utility case, where γp = γs =0, there is no need to keep separate track of

N j, but in the nonhomothetic case there is.

Both the primary and the service good in country j are produced according to constant returns production functions,

αp 1 αp j j j − Yp = θp Kp Hp (7)

j ¡ j ¢αs ¡ j ¢1 αs Ys = θs Ks Hs − . (8) ¡ ¢ ¡ ¢

In contrast, the technology for producing manufactured goods exhibits increasing returns to scale because of the presence of fixed costs. Specifically, every firm z has the

7 production function

αm 1 αm Ym (z)=max θmKm (z) Hm (z) − F, 0 . (9) − £ ¤

Here F>0 are the ﬁxed costs.

The ﬁrms in the manufacturing sector are monopolistic competitors. Firm z in country or region j sets its price qm (z) to maximize proﬁts

j j Π (z)=qm (z) Ym (z) r Km (z) w Hm (z) , (10) − −

j j j taking all of the other prices qp, qm (z0), qs, r , w as given. To do so, the ﬁrm solves the maximization problems of all the consumers to obtain the world demand function for its good

n j rw Ym (z)= Xm (z)+Xm (z) . (11) j=1 X

Here 1 1 η j j j j j j j − j βm r K + w H + qpγpN + qsγsN Xm (z)= ρ η (12) 1 ρ ρ(1− η) ¡ 1 ρ 1− ρ − ¢ qm (z) − Dw qm (z0) − dz0 ∆ hR i where η (1 ρ) − − − 1 η 1 η 1 ρ ρ − 1 η 1 η 1 η − 1 η 1 η 1− ρ 1 η j− − ∆ = βp− qp − + βm− qm (z0) − dz0 + βs − qs (13) w "µZD ¶ #

Given its choice of output, the ﬁrm chooses Km (z) and Hm (z) to minimize costs.

8 j j Let c (r ,w ,Ym (z)) be the solution to the cost minimization problem of ﬁrm z:

j αm j 1 αm j j w 1 r w − c r ,w ,Ym (z) = (Ym (z)+F ) . (14) θm αm 1 αm µ ¶ µ − ¶ ¡ ¢

Then we can write the proﬁts (10) of ﬁrm z as

ρ 1 1 ρ j 1 ρ j Π (z)=Aqm (z)− − C Aq (z)− − C F. (15) − −

Here we have expressed

1 1 ρ Ym (z)=Aqm (z)− − (16)

j j j c r ,w ,Ym (z) = C (Ym (z)+F ) (17) ¡ ¢ where A and Cj are the appropriate expressions derived from equations (11)—(14). Differen- tiating profits (15) with respect to qm (z) and setting the derivatives equal to zero yields the familiar Lerner condition for profit maximization

j qm (z)=C / ρ. (18)

Here the price elasticity of demand for good z is 1/ (1 ρ). It is straightforward to show − that this is the same result that we ﬁnd if we assume that ﬁrms set quantities rather than prices.

We determine the number of firms dw by allowing free entry and requiring that the profits of all firms are equal to zero. Using the Lerner condition (18), we can rewrite profits

9 (10) as

Π (z)=CjY w (z)/ρ CjY w (z) CjF. (19) m − m −

Setting this expression equal to zero, we obtain

ρ Y w (z)= F. (20) m 1 ρ −

j j j Definition 1. An equilibrium is a vector of prices qp, qm (z), qs, r ,andw , and quantities,

j j j j j j j j j j j w x , x (z), x , X , Xm (z), X , Y , K , H , Ym (z), Km (z), Hm (z), Y , K , H , z D , p m s p s p p p s s s ∈ j =1,...,n,rw, an interval of ﬁrms Dw,andameasureofﬁrms for each country or region,

Dj, j =1,...,n,rw, such that

j j j 1. Given the prices, the individual consumption plans xp, xm (z), xs, solve the utility max-

imization problem of consumer j (1)—(2);

2. The factor prices rj, wj, and the production plans for the primary and service good

satisfy the conditions for zero proﬁt and cost minimization

1 αp 1 αs j j j − j j − r = qpαpθp Hp / Kp = qsαsθs Hs / Ks (21)

j ¡ ¢j j αp ¡ ¢ j j αs w = qp (1 αp) θp K / H = qs (1 αs) θs K / H ; (22) − p p − s s ¡ ¢ ¡ ¢

3. Each manufacturing ﬁrm z in country or region j chooses price qm (z) to maximize

proﬁts (18). Given output Ym (z), it chooses inputs Km (z), Hm (z) to minimize costs;

4. Every ﬁrm z Dw earns zero proﬁts (20); ∈

10 5. The markets for goods clear,

n n n j j j j N xp = Xp = Yp (23) j=1 Ã j=1 ! j=1 n X n X X j j j w N xm (z) = Xm (z) = Ym (z) ,zD (24) j=1 Ã j=1 ! ∈ X X j j j j N xs = Xs = Ys ,j=1,...,n,rw; (25) ¡ ¢

6. The factor markets clear,

j j j Kp + Km (z) dz + Ks = K ,j=1,...,n,rw (26) j ZD j j j Hp + Hm (z) dz + Hs = H ,j=1,...,n,rw; (27) j ZD

7. The number of variety available for consumption is the number of varieties produced,

Dw = D1 D2 Dn Drw. (28) ∪ ∪ ···∪ ∪

If factor prices in the OECD and the rest of the world are equal, then all of the manufacturing ﬁrms are faced with symmetric problems. Consequently, they all set the same quantities and charge the same prices. Since there are 2 traded goods prices and 2 factors of production, we know that factor prices are, in fact, equal across regions if both regions produce both goods. This suggests a simple procedure for computing equilibrium common in trade models: We compute an integrated equilibrium model for the world economy. We then compute the production plan for each region. As long as both regions produce both goods, we are done with the computation. If one of the regions were to produce negative

11 amounts of one of the goods, we would be wrong to assume that factor prices were equal.

In this case we would have to go back and compute an equilibrium in which at least one of the countries specializes. (We are not very interested in these sorts of equilibria, however, because they do not correspond with world production patterns in 1961 and 1990.)

We solve the model for a world with two regions, the OECD and the rest of the world, eachofwhichismadeupofdiﬀerent countries. To see how the theory matches up with the data, however, it is essential that we be able to calculate intraindustry trade in manufactures within the OECD and between the OECD and the rest of the world.

To calculate intraindustry trade, we generalize the approach developed by Helpman and Krugman (1985). Let sj be the share of country or region j, j =1,...,n,rw in the world production of manufactures,

j j w s = Ym (z) dz Ym (z) dz = Ym / Ym . (29) Dj Dw Z . Z

j Let Xm be the total consumption of manufactures by country or region j,

j j Xm = Xm (z) dx. (30) w ZD

In the absence of trade barriers, the composition of consumption baskets of manufactured goods are the same in all countries and regions. Consequently, the imports of country j from the rest of the OECD are

M j = 1 srw sj Xj ,j=1,...,n (31) oe − − m ¡ ¢

12 The imports of the rest of the world from the OECD are

M rw =(1 srw) Xrw. (32) oe − m

j To obtain total trade within the OECD, we sum the expressions for Moe in (31)toobtain

n n oe j rw j 2 rw oe Moe = Moe = 1 s s (1 s ) Xm . (33) j=1 Ã − − j=1 − ! X X ¡ ¢ .

3. Calibration

In this section we describe the calibration of the model to a 1990 data set. We begin by assembling a benchmark data set for the OECD in 1990. Speciﬁcally, we aggregate ﬁgures on production and factor utilization for each of the 3 sectors in the model for the 23 countries listed in Table 1.

Table 2

Benchmark 1990 OECD Data Set

(Million U.S. Dollars) Primaries Manufactures Services Total oe Yi 668,993 3,659,294 12,141,293 16,469,580 oe Hi 228,208 2,883,736 8,543,962 11,755,906 oe Ki 440,785 775,558 3,497,331 4,713,674 oe Xi 861,634 3,466,653 12,141,293 16,469,580 Y oe Xoe -192,641192,641 00 i − i

The 1990 OECD data set is presented in Table 2. The ﬁgures for GDP in each of the 3 sectors are taken from OECD National Accounts. To obtain the factor inputs, we ﬁrst

13 obtain a labor compensation share for the OECD,

Hoe 23 LCj i=p,m,s i = j=1 . (34) oe 23 j j j j Yi (LC + FC + OS UP ) Pi=p,m,s j=1 P − P P Here LCj is the total labor compensation in country j; FCj is fixed capital consumption; OSj is operating surplus; and UPj is unincorporated profits. What this procedure does is to split indirect taxes and unincorporated profits, which is mostly returns to self-employed workers or family businesses, proportionally between returns to labor and returns to physical capital. We then proportionally adjust the labor compensations for each of the 3 sectors reported by OECD National Accounts so that their total yield the labor compensation implied by relation (34). Imports of primaries by the OECD from the rest of the world are taken from OECD Foreign Trade by Commodity. The number reported is that for net imports.

We also report results for an alternative calibration in which these imports are gross exports,

275,043 million U.S. dollars, rather than 192,641 million U.S. dollars. The results for this alternative calibration do not differ significantly from those reported in the next section. Net exports of manufacturing from the OECD to the rest of the world are set equal to imports of primaries to insure balanced trade. Notice though that, given product differentiation in manufacturing, the OECD both imports manufactured goods from the rest of the world and exports manufactured goods to it. The data on consumption by sector are obtained residu- ally. Notice that the concept of consumption in the model corresponds to consumption plus investment plus government spending in the national income accounts.

14 Population ﬁgures are taken from UN World Population Project.Theyare

N oe = 853.7,Nrw =4, 428.3, (35)

where the units are millions of people.

Table 3

Benchmark 1990 Rest of the World Data Set

(Million U.S. Dollars) Primaries Manufactures Services Total rw Yi 1,222,748 1,159,518 3,447,005 5,829,270 rw Xi 1,030,107 1,352,159 3,447,005 5,829,270 Y rw Xrw 192,641 -192,641 00 i − i

Total income in the rest of the world is taken from UN Yearbook of National Accounts

Statistics. In our model this ﬁgure is also equal to total consumption:

rw rw Yi = Xi =5, 829, 270. (36) i=p,m,s i=p,m,s X X

Table 3 presents a benchmark data set for the rest of the world. These numbers were also derived from the UN Yearbook of National Accounts Statistics using the following method- ology: We collect sectoral production data for any country for which such data is recorded for a year in the period 1984-1991. We use the sectoral production shares to impute sectoral production in 1990 by multiplying these shares by 1990 GDP. We are able to then impute sectoral production shares for the rest of the world and multiply these shares by total output in the rest of the world to impute sectoral outputs. The number of countries in the rest of the world for which we have sectoral output data is 103. The GDP of these countries in 1990

15 is 3,149,703 million U.S. dollars, which is 54.0 percent of total GDP in the rest of the world.

The value of ρ =0.833333 (=1/1.2) is chosen so that the markups charged by manufacturing ﬁrms over variable costs in the Lerner condition (18) is 20 percent. This is consistent with evidence presented by Morrison (1990). We normalize dw = 100. The choice

w j j j of any other value of d proportionally scales up or down F , xm(z),Xm(z), and Ym(z), but leaves the values of all other variables unchanged.

We calibrate the model by normalizing qp = qm(z)=qs = r = w =1and then calculating values of Krw and Hrw so that the benchmark data set is an equilibrium of the model. In numerical experiments in which we allow for nonhomothetic preferences, we can use the diﬀerent consumption shares in the OECD and the rest of the world to calibrate the utility parameters γp and γp. The calibration procedure yields a rest of the world that is more capital abundant than the OECD,

Krw Koe > (37) Hrw Hoe because the rest of the world needs to export the capital intensive good, primaries. This relative capital abundance is consistent with the limited data on sectoral labor shares in the UN Yearbook of National Accounts Statistics. It is also consistent with the evidence presented by Treﬂer (1993).

4. Numerical Experiments

In our numerical experiments, we introduce changes in the parameters of the model to simulate the world in 1961. The principal facts about 1961 that we incorporate into our

16 model are that the world was a much poorer place than in 1990 and that the distribution of income and consumption of manufactured goods was much more concentrated than in 1990.

Part of this concentration was reflected in the fact that the industrialized world, which we assume to be the OECD, consisted of fewer countries. Our model says that these differences will have effects on the direction and volume of trade.

In 1961 the OECD consisted of the 19 countries listed in Table 4. Notice the absence of Japan and the very large share of income generated by the United States.

Table 4

OECD in 1961 Country Share of Income (percent) Austria 0.5615 Belgium/Luxembourg 0.9354 Canada 2.5475 Denmark 0.5213 France 5.2392 Germany 7.2724 Greece 0.3778 Iceland 0.0205 Ireland 0.1569 Italy 3.4773 Netherlands 1.0276 Norway 0.4472 Portugal 0.2413 Spain 1.0300 Sweden 1.2098 Switzerland 0.8013 Turkey 0.6183 United Kingdom 6.0547 United States 41.7524

The world was much poorer in 1961 than it was in 1990 for two reasons: ﬁrst, endowments of factors were smaller; and, second, total factor productivities in the diﬀerent sectors were lower. We begin with total factor productivities by rescaling the constants θi

17 and the ﬁxed costs F in the production functions,

θp,1961 = θp,1990 (38)

29 29 θm,1961 = θm,1990/1.014 ,F1961 = F1990/1.014 (39)

29 θs,1961 = θs,1990/1.005 (40)

The yearly total factor productivity growth rates, 0, 0.014, and 0.005, are those derived by

Echevarria (1997) for the OECD.

Population ﬁgures, again taken from the UN World Population Project,are

oe rw N1961 = 536.0,N1961 =2, 545.0. (41)

oe oe rw rw We calibrate the four endowments, K1961, H1961, K1961, H1961, so the following four conditions are satisﬁed

oe oe i=p,m,s Yi,1990 /N1990 =2.4003 (42) ³ oe ´ oe Pi=p,m,s Yi,1961 /N1961

³ rw ´ rw Pi=p,m,s Yi,1990 /N1990 =2.0550 (43) ³ rw ´ rw Pi=p,m,s Yi,1961 /N1961 oe oe ³P ´ K1961 K1990 oe = oe (44) H1961 H1990 rw rw qp,1961(Yp,1961 Xp,1961) − rw =0.050285. (45) i=p,m,s qi,1961Yi,1961 P The yearly growth rates used in (42) and (43), 0.030653 for the OECD and 0.025148 for the OECD, are derived from various issues of the World Bank World Development Report.

18 Unfortunately, these growth rates are calculated for real GDP data that are chained in a complicated way, rather than based on a ﬁxed base year’s prices. An alternative is to compute growth rates for real GDP per capita based on 1961 prices, with

oe oe qi,1961Y /N i=p,m,s i,1990 1990 =2.473882, (46) q Y oe /N oe Pi=p,m,s i,1961 i,1961 1961 P for example. Yet another possibility would be to take the growth data for the OECD in (42) as given, but to replace the growth data for the rest of the world (43) with the requirement that oe qi,1961Y i=p,m,s i,1961 =2.985134, (47) q Y rw Pi=p,m,s i,1961 i,1961 P whichsaysthattheratioofOECDtorestoftheworldGDP,at1961 prices, should equal that observed in the data. Results for numerical experiments with this alternative calibration are reported in the next section. They do not diﬀer signiﬁcantly from those reported here.

Requirement (44), that the capital/labor ratio in the OECD stay fixed has no significant effect on our results given the other requirements that we are imposing. Requirement

(45) says that net exports of primaries from the rest of the world to the OECD should equal their observed value in 1961, taken in this case from GATT. As we have already explained it is equally possible to calibrate the model to reproduce gross exports as a faction of GDP, which were 0.081632 rather than 0.050285. The next section also presents an experiment in which we require, rather than satisfying (45), that

rw rw K1961 K1990 rw = rw , (48) H1961 H1990

19 Although the results for this alternative calibration are slightly more favorable to the new trade theory, they also imply that the rest of the world should have been a net importer of primaries in 1961, a result drastically at odds with the data.

Table 5

Results for Base Case Calibration 19611990 Change Data OECD-OECDTrade/OECDIncome 0.0531 0.1123 111.5% OECD-OECD Trade / OECD-RW Trade 0.8435 1.5786 87.1% OECD Manf Con / OECD Income 0.2779 0.2105 -24.3%

1. γp =0, γs =0, η =0 OECD-OECDTrade/OECDIncome 0.1070 0.1357 26.8% OECD-OECD Trade / OECD-RW Trade 0.8927 1.1685 30.9% OECD Manf Con / OECD Income 0.2105 0.2105 0.0%

2. γp = 169.5, γs = 314.7, η =0 OECD-OECDTrade/OECDIncome− 0.1023 0.1322 29.2% OECD-OECD Trade / OECD-RW Trade 0.7392 1.0603 43.4% OECD Manf Con / OECD Income 0.2123 0.2105 -0.8%

3. γp = 169.5, γs = 314.7, η =0.4372 OECD-OECDTrade/OECDIncome− 0.0625 0.1322 111.5% OECD-OECD Trade / OECD-RW Trade 0.7384 1.0603 43.6% OECD Manf Con / OECD Income 0.1253 0.2105 68.0%

4. γp = 169.5, γs = 314.7, η = 1.2150 OECD-OECDTrade/OECDIncome− − 0.1323 0.1322 -0.1% OECD-OECD Trade / OECD-RW Trade 0.7398 1.0603 43.3% OECD Manf Con / OECD Income 0.2779 0.2105 -24.3%

Table 5 reports the results of some numerical experiments with our model. We focus

ﬁrst on the base line experiment in which γp =0, γs =0, η =0. It is this experiment that is the best test of the Helpman-Krugman explanation of the expansion of trade volume. Notice that, although trade between OECD countries as a fraction of income does expand by 26.8 percent, this increase is far short of the 111.5 percent increase in the data. Notice too that trade within the OECD increases only by 30.9 percent compared to OECD trade with the rest of the world, rather than increasing by 87.1 percent as in the data. The Helpman-Krugman

20 explanation of the increase in trade volumes, embodied in this experiment falls well short of accounting for the facts.

Let us now focus on the experiments in which utility is nonhomothetic. Notice that, when we calibrate the parameters γp and γs to match the consumption shares in Tables

2 and 3, setting γ = 169.5, γ = 314.7, we obtain parameters that are consistent with p − s other evidence that it is services that have the highest income elasticity of demand, followed by manufactures, which are in turn followed by agriculture. As we can see, Markusen’s

(1986) story does indeed go a long way in accounting for the increase in OECD-OECD trade compared to OECD trade with the rest of the world, accounting for almost half of the observed increase.

The next experiment, in which γ = 169.5, γ = 314.7,andη =0.4372, shows that, p − s if we introduce price elasticities of demand that differ from 1 by letting η differ from 0, the model is indeed flexible enough to account for the increase of OECD-OECD trade compared to OECD income. The value of η that we use is very high, however: Stockman and Tesar

(1995), for example, estimate η to be 1.27. Notice too that this parameterization results in − a huge increase in the share of manufactures in consumption in the OECD over the period

1961-1990 as their relative price falls because of technological progress. This huge increase in share is very much at odds with the decline observed in the data.

The ﬁnal experiment shows that, for a reasonable value of η, η = 1.2150, the model − is capable of matching the decline in consumption share of manufactures while preserving the explanation for the expansion of OECD-OECD trade relative to OECD trade with the rest of the world. Notice that this experiment predicts that OECD-OECD trade should have fallen slightly as a fraction of OECD income, a prediction drastically at odds with the data.

21 The ﬁnal numerical experiment is the only one consistent with the observed diﬀerences in the composition of output between the OECD and the rest of the world in 1990 and the observed change in the composition of output in the OECD between 1961 and 1990. In this numerical experiment the new trade theory fails to account for any of the increase in the ratio of OECD-OECD trade to OECD income. The theory does, however, account for half of the increase in ratio of OECD-OECD trade to OECD-RW trade.

Some simple calculations show why we should not have expected the new trade theory to simultaneously account for both the sharp increase in trade over the period 1961-1990 and the decline in the importance of manufacturing in ﬁnal demand. Neglecting trade in manufactures with the rest of the world, we can approximate OECD-OECD trade in manufactures with the formula

M oe n Xoe oe (1 (sj)2) m . (49) Y oe ≈ Y oe − j=1 X

In the data, the index of size distribution of national incomes in the OECD, (1 n (sj)2), − j=1 goes from 0.6634 in 1961 to 0.8272 in 1990, producing the increase in trade toP income emphasized by Helpman and Krugman (1985). Itisherethatallowingthemembershipofthe

OECD to increase over time produces results favorable to the theory. The ratio of man-

oe oe ufacturing GDP to total GDP in the OECD, Xm /Y , falls, however, from 0.2779 in 1961 to 0.2105 in 1990. Notice that these two changes almost exactly cancel each other out, producing the prediction of no increase in the ratio of OECD-OECD trade to OECD income trade in the ﬁnal experiment.

22 5. Sensitivity Analysis

This section reports the results of numerical experiments of models in which we employ alternative calibration methodologies. First, we report the results of experiments where, rather than calibrating the utility parameters γp and γs to match observed consumption shares, we set them arbitrarily to γ = 200 and γ =1600. These parameter values p − s make primaries more inferior and services more superior than in our base case calibration.

One defense for this alternative speciﬁcation is that the data in Table 3, upon which our calibration of the parameters γp and γs is based, are probably the least reliable numbers in our benchmark data set.

Table 6

Results for Alternative Speciﬁcation of Nonhomotheticity 19611990 Change Data OECD-OECD Trade / OECD Income 0.0531 0.1123 111.5% OECD-OECD Trade / OECD-RW Trade 0.8435 1.5786 87.1% OECD Manf Con / OECD Income 0.2779 0.2105 -24.3%

1. γp = 200, γs =1600, η =0 OECD-OECD− Trade / OECD Income 0.0788 0.1093 38.7% OECD-OECD Trade / OECD-RW Trade 0.3263 0.6074 86.1% OECD Manf Con / OECD Income 0.2269 0.2105 -7.2%

2. γp = 200, γs =1600, η =0.4170 OECD-OECD− Trade / OECD Income 0.05170.1093 111.5% OECD-OECD Trade / OECD-RW Trade 0.3259 0.6074 86.9% OECD Manf Con / OECD Income 0.1449 0.2105 45.3%

3. γp = 200, γs =1600, η = 0.7416 OECD-OECD− Trade / OECD− Income 0.0957 0.1093 14.2% OECD-OECD Trade / OECD-RW Trade 0.3265 0.6074 86.0% OECD Manf Con / OECD Income 0.2779 0.2105 -24.3%

Notice in Table 6 that the results for both the ratio of OECD-OECD trade to income and the ratio of OECD-OECD trade to OECD-RW trade improve signiﬁcantly. Even so, in the third experiment, where η is calibrated to match the observed decline in the consumption

23 share of manufactures, the model is only able to replicate a small fraction of the observed increase in the ratio of OECD-OECD trade to OECD income. Nevertheless, no matter what the value of η, Markusen’s (1986) story based on inferiority of primaries can account for the observed increase in the ratio of OECD-OECD trade to OECD trade with the rest of the world if utility is suﬃciently nonhomothetic.

Table 7

Results for Gross Imports Calibration 19611990 Change Data OECD-OECDTrade/OECDIncome 0.0531 0.1123 111.5% OECD-OECD Trade / OECD-RW Trade 0.8435 1.5786 87.1% OECD Manf Con / OECD Income 0.2674 0.2055 -23.1%

1. γp =0, γs =0, η =0 OECD-OECDTrade/OECDIncome 0.1081 0.1357 25.5% OECD-OECD Trade / OECD-RW Trade 0.9029 1.1685 29.4% OECD Manf Con / OECD Income 0.2055 0.2055 0.0%

2. γp = 134.6, γs = 418.0, η =0 OECD-OECDTrade/OECDIncome− 0.0994 0.1291 29.8% OECD-OECD Trade / OECD-RW Trade 0.6440 0.9759 51.5% OECD Manf Con / OECD Income 0.2090 0.2055 -1.7%

3. γp = 134.6, γs = 418.0, η =0.4440 OECD-OECDTrade/OECDIncome− 0.06100.1291 111.5% OECD-OECD Trade / OECD-RW Trade 0.6430 0.9759 51.8% OECD Manf Con / OECD Income 0.1212 0.2055 69.6%

4. γp = 134.6, γs = 418.0, η = 0.9879 OECD-OECDTrade/OECDIncome− − 0.1250 0.1291 3.3% OECD-OECD Trade / OECD-RW Trade 0.6448 0.9759 51.3% OECD Manf Con / OECD Income 0.2674 0.2055 -23.1%

Table 7 reports the results of numerical experiments of a model in which imports of primary goods by the OECD from the rest of the world are identiﬁed with gross exports, rather than with net exports as in the base case calibration. Notice that, given our calibration procedure, the fraction of income spent on manufactures in the OECD in 1990 changes from

0.2105 to 0.2055.

24 Table 8 reports the results of a set of numerical experiments for the calibration in which growth in the endowments of the rest of the world between 1961 and 1990 are calibrated to replicate the observed ratio of OECD income to income in the rest of the world in 1961 (47), rather than to replicate the observed growth rate (43). Similar calculations, not reported here, show that imposing growth rates based on 1961 prices (46) do not signiﬁcantly aﬀect the results.

Table 8

Results for Alternative RW Growth Calibration 19611990 Change Data OECD-OECDTrade/OECDIncome 0.0531 0.1123 111.5% OECD-OECD Trade / OECD-RW Trade 0.8435 1.5786 87.1% OECD Manf Con / OECD Income 0.2779 0.2105 -24.3%

1. γp =0, γs =0, η =0 OECD-OECDTrade/OECDIncome 0.1094 0.1357 24.0% OECD-OECD Trade / OECD-RW Trade 0.99011.1685 18.0% OECD Manf Con / OECD Income 0.2105 0.2105 0.0%

2. γp = 169.5, γs = 314.7, η =0 OECD-OECDTrade/OECDIncome− 0.1049 0.1322 26.0% OECD-OECD Trade / OECD-RW Trade 0.8182 1.0603 29.6% OECD Manf Con / OECD Income 0.2123 0.2105 -0.8%

3. γp = 169.5, γs = 314.7, η =0.4437 OECD-OECDTrade/OECDIncome− 0.0625 0.1322 111.5% OECD-OECD Trade / OECD-RW Trade 0.8152 1.0603 30.1% OECD Manf Con / OECD Income 0.1226 0.2105 71.7%

4. γp = 169.5, γs = 314.7, η = 1.1907 OECD-OECDTrade/OECDIncome− − 0.1359 0.1322 -2.7% OECD-OECD Trade / OECD-RW Trade 0.8205 1.0603 29.2% OECD Manf Con / OECD Income 0.2779 0.2105 -24.3%

The ﬁnal set of results reported are for numerical experiments of a model in which endowments in the rest of the world in 1961 are required to have the same capital/labor ratioastheydoin1990 (48) rather than to generate the observed exports of primaries to the

OECD (45).

25 Table 9

Results for Alternative Endowment Calibration 19611990 Change Data OECD-OECD Trade / OECD Income 0.0531 0.1123 111.5% OECD-OECD Trade / OECD-RW Trade 0.8435 1.5786 87.1% OECD Manf Con / OECD Income 0.2779 0.2105 -24.3% RW Prim Exp / RW Income 0.0503 0.0330 -34.3%

1. γp =0, γs =0, η =0 OECD-OECD Trade / OECD Income 0.1077 0.1357 26.0% OECD-OECD Trade / OECD-RW Trade 0.8948 1.1685 30.6% OECD Manf Con / OECD Income 0.2105 0.2105 0.0% RW Prim Exp / RW Income 0.0565 0.0330 -41.5%

2. γp = 169.5, γs = 314.7, η =0 OECD-OECD− Trade / OECD Income 0.0920 0.1322 29.8% OECD-OECD Trade / OECD-RW Trade 0.71811.0603 47.7% OECD Manf Con / OECD Income 0.2123 0.2105 -0.8% RW Prim Exp / RW Income -0.0407 0.0330 -

3. γp = 169.5, γs = 314.7, η =0.3555 OECD-OECD− Trade / OECD Income 0.0625 0.1322 111.5% OECD-OECD Trade / OECD-RW Trade 0.7166 1.0603 48.0% OECD Manf Con / OECD Income 0.1497 0.2105 40.6% RW Prim Exp / RW Income -0.0497 0.0330 -

4. γp = 169.5, γs = 314.7, η = 1.2361 OECD-OECD− Trade / OECD− Income 0.1230 0.1322 3.3% OECD-OECD Trade / OECD-RW Trade 0.7199 1.0603 47.3% OECD Manf Con / OECD Income 0.2779 0.2105 -24.3% RW Prim Exp / RW Income -0.0311 0.0330 -

Notice that this calibration results in the rest of the world importing primary goods form the

OECD in 1961.

6. Some Not So Recent Trade Facts

Although the three facts reported in the introduction do indeed characterize post

WorldWarIItradedata,theydonotcharacterizedatabeforethen. Thenewtradetheorists of the 1980s may have gone too far in focusing on a limited amount of data. The historical data cast doubt on the explanations of the facts posited by the new trade theory.

26 The high rates of growth in foreign trade and the steady increase in the ratio of trade to income observed after the 1950s, also characterized the trends in the foreign trade statistics during the nineteenth century. In fact, as reported by Kuznets (1967), between 1800 and

1913, per capita world trade grew at an average rate of 33.0 percent per decade, whereas per capita world income did it at an average rate of 7.3 percent. As a result, during the period, the ratio of trade to income increased to over 11 times its initial level. Since estimates for

1913showthattheratioofworldexportstoworldincomewasabout17percent,in1800 it must have been below 2 percent. The inter war period, however, resulted in a dramatic reduction in trade, not only as a fraction of world income, but also in absolute terms. This reduction was particularly intense during the Great Depression years. United Nations data show that by 1950 the ratio of world trade to world income had fallen to less than 8 percent.

By 1990, this ratio had risen to slightly more than 15 percent, still not at the level achieved in 1913.

Data for the United States show a similar pattern. Starting in the 1960s, as Figure

4 reports, there was a signiﬁcant increase in the ratio of trade to income but only to reach, in 1990, a level similar to those seen during the second half of the nineteenth century. (The data in this ﬁgure calculate trade as exports plus imports.)

Finally, looking at directions of trade, Woytinsky and Woytinsky (1955) report that

Europe dominated world trade during the nineteenth century. In 1913 the ratio of intra

European trade to world trade was 40 percent. By 1938, however, it had fall to 29 percent, and in 1953 it was 22 percent. During the next thirty years this ratio increased steadily until reaching 38 percent in 1990, a value similar to the one seen in 1913.

27 7. Intermediate Goods?

Our model does not include intermediate goods. Yet much of the increase in trade has been in intermediate goods. (See, for example, Feenstra, 1998.) Could introducing intermediate goods be a way of rescuing the new trade theory? As Figure 5 shows, in the United

States and in Mexico the utilization of intermediate goods has declined sharply over the past

20 years compared to GDP. Looking at Input-Output Matrices conﬁrms that the decline in the importance of intermediate goods reﬂects the decline in importance of manufactures. In- termediate goods are disproportionately used by and produced by the manufacturing sector.

In the United States in 1987, although the manufacturing sector accounted for 21.5 percent of GDP, it accounted for 40.0 percent of the use of intermediate goods and 38.1 percent of their production. The Mexican data are similar: In Mexico in 1985, although the manufacturing sector accounted for 23.4 percent of GDP, it accounted for 52.3 percent of the use of intermediate goods and 41.7 percent of their production. Adding intermediate goods to the models would complicate the data analysis because of the lack of comparable input-output matrices across countries and time. It would still leave us with the same mystery in relation tothenewtradetheory: Thegoodsthatarebeingtradedmoreandmoreovertimearethe same goods whose importance is falling in relationship to domestic production.

8. Policy?

The post war years have seen substantial steps towards global trade liberalization.

Could it be changes in policy rather than the features emphasized in the new trade theory that have been responsible for the dramatic increase in the ratio of trade to income? We can provide a preliminary answer to this question and highlight the issues at stake with a simple

28 version of the new trade theory model used in this paper.

Consider a model with only one sector, the manufacturing sector, and one set of countries that engage in international trade, the OECD. Suppose that each of the n countries in the OECD imposes trade barriers on imports from the other countries in the form of a uniform ad valorem tariff, τ. In contrast to our earlier analysis, we assume that all of the countries in the OECD are identical in terms of size, because trade barriers would affect countries of different size differently. In this model, each country would produce the fraction

1/n of the world’s varieties of goods. Let Xd be the amount of each variety consumed domestically and Xf theamountconsumedineachofthen 1 foreign countries. Symmetry − and the ﬁrst order conditions for utility maximization imply that

Xd 1 1 ρ =(1+τ) − . (50) Xf

Market clearing implies that

Xd +(n 1)Xf = Y. (51) −

Combining these conditions, we can obtain an expression for the ratio of exports to income:

(n 1)Xf n 1 − = − 1 . (52) Y n 1+(1+τ) 1 ρ − −

To replicate index of size distribution of national incomes in the OECD, (1 n (sj)2), − j=1 with symmetric countries where sj =1/n, we can let the number of countries takeP on non integer values. In 1961, (1 n (sj)2)=0.6634, implying that n =2.97. In 1990, − j=1 (1 n (sj)2)=0.8272, implyingP that n =5.79. As we have seen, however, the change − j=1 P 29 in the size distribution of national incomes is almost exactly canceled out by the decline in the importance of the manufacturing sector. Consequently, we ﬁx n =5.79 and ask whether changes in trade barriers as represented by τ can account for the more than doubling of the ratio of trade to income. The answer to this question obviously depends on how much trade barriers have fallen and on the elasticity of substitution between varieties, 1/(1 ρ). − Calculations for a wide variety of parameters are presented in Figure 6. What we need is a large fall in trade barriers, accompanied by a large elasticity of substitution. If ρ =1/1.1, for example, implying an elasticity of substitution of 11, a fall in τ from 0.25 to 0.05 implies that the ratio of trade to income increases by a factor of 2.5. If ρ =1/1.2, however, implying an elasticity of substitution of 6, we need a larger fall in τ, say from about 0.50 to 0.05 to produce the same sort of increase in the ratio of trade to income.

Yi (1999) argues that, since average tariﬀ rates have fallen from about 15percentin

1960to5percentin1990, incorporating policy changes into the new trade theory cannot account for the increase in trade unless we assume very high elasticities of substitution in varieties. He presents a model in which it is increases in international vertical integration, induced by changes in trade policy, that account for the increase in the ratio of trade to product. It must be pointed out, however, that Harrigan (1993)showsthateveninthe

OECD in 1983 non tariﬀ trade barriers were far higher than tariﬀ barriers, in fact more than

8 times higher. Hummels (1999) has also identiﬁed a large number of trade barriers and has shown that their presence does a good job in accounting for observed trade patterns. To the extent that these trade barriers have fallen signiﬁcantly, a version of the new trade model that emphasizes trade policy seems capable of explaining large increases in the ratio of trade to income.

30 Data sources

We report the sources for all data used in the paper ordered as they are presented.

Indices of output and exports for 1950-1990 in Figure 1:

United Nations, Statistical Yearbook, New York, various issues.

World exports and world GDP for 1950, 1970 and 1990:

United Nations, Trends in International Distribution of Gross World Product, Special Issue,

National Account Statistics,NewYork,1993.

Trade within the OECD and OECD trade with the rest of the world for 1961-1990 in Figures 2 and 3:

Organization for Economic Cooperation and Development, Foreign Trade by Commodities, volumes 1 and 2, Paris, various issues.

Grubel-Lloyd indices for 1990:

Organization for Economic Cooperation and Development, Foreign Trade by Commodities, volumes 1 and 2, Paris, 1993.

United Nations, International Trade Statistics Yearbook, New York, 1993.

GDP for each OECD country in Tables 1 and 4 and sectoral GDP and labor and capital for the OECD in Table 2:

Organization for Economic Cooperation and Development, National Accounts, Paris, various issues.

Net and gross imports of primaries by the OECD from the rest of the world for

1990 in Table 2:

31 Organization for Economic Cooperation and Development, Foreign Trade by Commodities, volume 1,Paris,1993.

Population for the OECD and the rest of the world in 1990 and 1961:

United Nations, World Population Project,NewYork,variousissues.

Aggregate and sectoral GDP for the rest of the world in 1990 in Table 3:

United Nations, Yearbook of National Account Statistics,NewYork,1993.

IncomepercapitagrowthratesfortheOECDandtherestoftheworldfor

1961-1990:

World Bank, World Development Report, Washington, various issues.

Net exports of primaries from the rest of the world to the OECD in 1961:

General Agreement on TariﬀsandTrade,International Trade,Geneva,1963.

International Monetary Fund, Direction of Trade Annual, Washington, 1965.

Exports, imports, and GNP for the United States for 1870-1990 in Figure 4:

Department of Commerce, Bureau of the Census, Historical Statistics of the United States:

Colonial Times to 1970, Washington, 1975.

United Nations, National Account Statistics, Main Aggregates and Detailed Tables,NewYork, various issues.

Historical data on world trade:

S. Kuznets, “Quantitative Aspects of the Economic Growth of Nations: X-Levels and Struc- ture of Foreign Trade: Long-term Trends,” Economic Development and Cultural Change,

Part II, 1967.

32 United Nations, Trends in International Distribution of Gross World Product, Special Issue,

National Account Statistics,NewYork,1993.

Historical data on trade within Europe:

W. S. Woytinsky and E. S. Woytinsky, World Commerce and Governments: Trends and

Outlook, The Twentieth Century Fund, New York, 1955.

International Monetary Fund, Direction of Trade Statistics Yearbook, Washington, various issues.

Intermediate inputs and GDP in the United States and Mexico in Figure 5:

Instituto Nacional de Estadística, Geograﬁa e Informática, Anuario Estadístico de los Estados

Unidos Mexicanos, Aquascalientes, various issues.

Department of Commerce, Bureau of Economic Analysis, “Improved Estimates of Gross

Product by Industry 1947-98,” Survey of Current Business, 81 (2000). The data are available on the Bureau’s website, http://www.bea.doc.gov/bea.

Input-output matrices for the United States and Mexico:

Instituto Nacional de Estadística, Geograﬁa e Informática, Matriz de Insumo Producto Ac- tualizada a 1985,Mexico,D.F.,1990.

Department of Commerce, Bureau of Economic Analysis, “Input-Output Accounts of the

U.S. Economy, 1987,” Survey of Current Business, 74 (1994). The data are available on the

Bureau’s website, http://www.bea.doc.gov/bea.

33 References

Deardorﬀ,A.V.(1984), “Testing Trade Theories and Predicting Trade Flows,” in R. W. Jones

and P. Kenen, editors, Handbook of International Economics.Volume1.Amsterdam:

North-Holland, 467-517.

Dixit, A. K. and J. E. Stiglitz (1977), “Monopolistic Competition and Optimum Product

Diversity,” American Economic Review, 67, 297-308.

Echevarria, C. (1997), “Changes in Sectoral Composition Associated with Economic Growth,”

International Economic Review, 38, 431—452.

Feenstra,R.C.(1998), “Integration of Trade and Disintegration of Production in the Global

Economy,” Journal of Economic Perspectives, 12,31-50.

Harrigan, J. (1993), “OECD Imports and Trade Barriers in 1983,” Journal of International

Economics, 35,91-112.

Haveman, J. and D. Hummels (1999), “Alternative Hypotheses and the Volume of Trade:

Evidence on the Extent of Specialization,” Purdue University, unpublished manuscript.

Helpman, E. (1981), “International Trade in the Presence of Product Diﬀerentiation, Economies

of Scale, and Monopolistic Competition: A Chamberlin-Heckscher-Ohlin Approach,”

Journal of International Economics, 11, 305-340.

Helpman, E. (1987), “Imperfect Competition and International Trade: Evidence from Four-

teen Industrial Countries,” Journal of the Japanese and International Economies, 1,

62-81.

Helpman, E. and P. Krugman (1985), Market Structure and International Trade.MITPress.

34 Hummels, D. (1999), “Toward a Geography of Trade Costs,” University of Chicago, unpub-

lished manuscript.

Hummels, D. and J. Levinsohn (1995), “Monopolistic Competition and International Trade:

Reconsidering the Evidence,” Quarterly Journal of Economics, 110, 799-836.

Hunter, L. (1991), “The Contribution of Nonhomothetic Preferences to Trade,” Journal of

International Economics, 30, 345-358.

Krugman, P. (1979), “Increasing Returns, Monopolistic Competition, and International Trade,”

Journal of International Economics, 9, 469-479.

Kuznets, S. (1967), “Quantitative Aspects of the Economic Growth of Nations: X-Levels and

Structure of Foreign Trade: Long-term Trends,” Economic Development and Cultural

Change,PartII.

Lancaster, K. (1980), “Intra-Industry Trade Under Perfect Imperfect Competition,” Journal

of International Economics, 10, 151-175.

Markusen,J.R.(1986), “Explaining the Value of Trade: An Eclectic Approach,” American

Economic Review, 76, 1002—1011.

Markusen, J. R., J. R. Melvin, W. M. Kaempfer, and K. Maskus (1995), International Trade:

Theory and Evidence. McGraw-Hill.Morrison, C. (1990), “Productive and Financial

Performance in U.S. Manufacturing Industries: An Integrated Structural Approach,”

Southern Economic Journal, 60, 376-392.

Spence, A. M. (1976), “Product Selection, Fixed Costs, and Monopolistic Competition,”

Review of Economic Studies, 43,217-235.

35 Stockman, A. C. and L. L. Tesar (1995), “Tastes and Technology in a Two-Country Model

of the Business Cycle: Explaining International Comovements,” American Economic

Review, 85, 168-185.

Treﬂer,D.(1993), “International Factor Price Diﬀerences: Leontief was Right!” Journal of

Political Economy, 101,961-987.

Woytinsky, W. S., and E. S. Woytinsky (1955), World Commerce and Governments: Trends

and Outlook. The Twentieth Century Fund.

Yi,K-M.(1999), “Can Vertical Specialization Explain the Growth of World Trade,” Federal

Reserve Bank of New York, unpublished manuscript.

36 1100

1000

900

800

700 exports

600

500

400

GDP 300

200

100 1950 1955 1960 1965 1970 1975 1980 1985 1990 year

Figure 1: World Exports and World GDP (1950=100)

37 1.80

1.60

1.40

1.20

1.00

0.80

0.60

0.40

0.20

0.00 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 year

Figure 2: OECD-OECD Trade / OECD-RW Trade

38 0.12

0.10

0.08

0.06

0.04

0.02

0.00 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 year

Figure 3: OECD-OECD Trade / OECD Income

39 0.20

0.18

0.16

0.14

0.12

0.10

0.08

0.06

0.04

0.02

0.00 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 year

Figure 4: U.S. Total Trade / U.S. GNP

40 1.10

1.00

0.90

0.80 United States

0.70 Mexico

0.60

0.50

0.40 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 year

Figure 5: Intermediate Inputs / GDP in the United States and Mexico

41 0.90

0.80

0.70

rho=1/2.0 0.60

rho=1/1.5 0.50

0.40

rho=1/1.2 0.30

0.20 rho=1/1.1 0.10

0.00 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 tariff rate

Figure 6: World Trade / World Income